Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: Ph.D. Dissertation
- Language: Farsi
- Document No: 43900 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Bagheri Shouraki, Saeed
- Abstract:
- Wireless Sensor Networks (WSNs) attracted lots of researchers in recent years, these sensors are small and have limited computational resources and are not expensive in comparision with conventional sensors. These sensors can sense, measure, gathering information from environment and send them to user, they are typically characterized by limited communication capabilities due to tight energy and bandwidth constraints. As a result, WSNs have inspired a resurgence in research on decentralized algorithms, structure of these networks are variable and dynamic, sensors may deleted from the network or added to the network, so using models which don’t require precies specifications of environment is desirable, by using these models, the same sensors, without reprogramming, could be used in different environmets, one of the methods which don’t require precies specifications of environment and could be used for different goals in WSNs is distributed machine learning.In many real scenarios in WSNs, we have limited or unprecies information about the environment which couldn’t be used in current models for machine learning in WSNs.
In this thesis a new method for machine learning in WSNs is proposed which could using supervision information, this method is similar to Q-Learning and has a low communication and computational overhead, this method is analysed theoretically and the positive effect of supervision information in all steps of learning is showed, furthermore a distributed machine learning method for adjusting the transmission power of sensors in sparse WSNs is proposed which doesn’t require any location information, this method is analysed theoretically in terms of energy consumption and computational complexity and communication complexity, furthermore to better use of supervision information and specifying the bounds of supervision information, which these informations are usable, a new theoretical bound between sensing range and communication range of sensors is proposed and the previous bound is improved. All proposed algorithms and methods in this thesis are tested through simulation - Keywords:
- Wireless Sensor Network ; Machine Learning ; Reinforcement Learning ; Side Information ; Energy Consumption ; Transmission Power Adjustment ; Communication Range ; Sensing Range
Digital Object List-
محتواي پايان نامه - view
Bookmark